DE102017203082B4 - Parallel imaging with improved reference data set to determine the weighting matrix - Google Patents

Abstract

A method of generating MR image data of a parallel imaging subject, comprising the steps of:
Picking up a reference data set of the examination subject in which at least a subsection of the associated raw data space is completely filled with raw data according to the Nyquist condition,
Recording an imaging data set for the generation of the MR image data on the basis of spin-echo-based signals, in which the associated raw data space is not completely recorded according to the Nyquist condition,
Calculating a weighting matrix on the basis of the reference data record with which raw data points of the imaging data record not recorded in the parallel imaging are determined,
Calculating the unrecorded raw data points of the imaging data set using the calculated weighting matrix, wherein the reference data set is taken with a spin echo based reference imaging sequence without segmentation within an imaging layer and without temporal overlap with the acquisition of the imaging data set and a resolution is used in the readout direction which is at least a factor of 4 less than when recording the imaging data set, wherein in the preparation of the reference data set RF refocusing pulses are used after a first RF refocusing pulse to generate the spin echoes having a smaller tilt angle than the RF Refocusing pulses after the first refocusing pulse used in the creation of the imaging data set.

Description

The present invention relates to a method for generating MR image data of a subject with parallel imaging and the associated MR system. Furthermore, a computer program product is provided and an electronically readable data carrier.

Cylindrical geometry MR systems have reduced homogeneity of the basic magnetic field B 0 and reduced gradient linearity at the edges of the field of view in the bore direction, commonly referred to as the z-axis. For coronary and sagittal recordings, this results in compressed areas in the reconstructed high signal intensity images in the areas at the edge of the bore, because the inhomogeneity of the gradient field and / or the polarization field B0 mimics multiple areas of an examination subject on a relatively small image area the signal intensity increases in this image area, in turn reduced in adjacent areas. This means that due to the incorrect spatial coding, pixels with increased as well as reduced signal intensity exist. These areas are usually located at the edge of the field of view in the z-direction, also called field of view (FOV). In addition, in imaging sequences using the parallel imaging technique, such as GRAPPA, artifacts often occur at integer multiples of the FOVz / R position, where R is the acceleration factor. In parallel imaging, the missing unmeasured raw data areas are calculated by means of several receiving channels and their different coil sensitivities. The coil sensitivities are determined by so-called reference measurements or calibration measurements. However, this requires a good reference or calibration data set. If this is faulty, then the MR images calculated with the parallel imaging are also faulty. The reason for the parallel imaging artifacts also lies in the fact that the imaging algorithms are usually unable to handle very localized areas of very high signal intensity which, as explained above, may occur at the periphery of the MR system components. This leads in particular in the parallel imaging to the incomplete suppression of so-called aliasing artifacts in these areas.

This phenomenon is particularly prominent in spin-echo sequences, where dephasing due to B0 inhomogeneities or gradient inhomogeneities plays a lesser role and does not lead to signal extinction as in gradient echo sequences. In addition, the strength and shape of the aforementioned artifacts are dependent on the acquisition parameters (such as gradient amplitudes, RF pulse bandwidth, reception bandwidth and image resolution).

Another constellation that can lead to the inclusion of a faulty calibration data record is the segmented recording of the calibration data. This means that the required k-space data is not recorded consecutively in a short time sequence, but segmented over a longer period of time. This is particularly the case when the data is obtained as part of the acquisition of the imaging data. Physiological effects such as blood flow, liquor pulsation, heartbeat and respiration then lead to the segmented recorded data being inconsistent with each other, for example as regards their relative phase relationship to one another.

One way to reduce these artifacts is to average MR signals, where the MR signals are picked up two or more times, for example twice, i. once recording the straight k-space lines and once recording the odd k-space lines. The central k-space lines of both images are then combined and used to calculate the reference and calibration data sets. By such a method, the artifacts in the final composite image cancel, but at least the measurement time doubles, which is a significant drawback.

US 2010/0303 320 A1 describes a method for fat and water separation using a fast spin-echo imaging sequence, wherein reference recordings are performed with reduced resolution compared to the recordings for fat and water separation.

US 2009/0 093 709 A1 discloses a method for noise suppression in parallel imaging wherein relatively low resolution MR data is acquired for generating the weighting matrix separate from the actual imaging.

It is an object of the invention to reduce the above-mentioned drawbacks in order to improve the parallel imaging so that there are fewer artifacts in the reconstructed image data with simultaneous reduction of the recording time. This object is solved by the features of the independent claims.

According to a first aspect, there is provided a method of generating MR image data of a parallel imaging subject. A reference data record of the examination subject is recorded, in which at least one subsection of the associated raw data space is completely filled with raw data according to the Nyquist condition. Furthermore, an imaging data set for the generation of the MR image data is recorded on the basis of spin-echo-based signals, wherein the imaging data record does not completely include the associated raw data space according to the Nyquist condition. Likewise, a weighting matrix is calculated on the basis of the reference data set with which raw data points of the imaging data record not recorded during parallel imaging are determined. The unrecorded raw data points of the imaging data set are calculated using the calculated weighting matrix. Here, the reference data set is acquired with a spin-echo-based reference imaging sequence without segmentation of the acquisition data within an imaging layer and without temporal overlap with the acquisition of the imaging data set, and the resolution in the readout direction is at least a factor of four lower than when recording the imaging data set ,

By using a time-separated reference data set, which does not overlap with the time the imaging data record is taken, the resolution in the read-out direction can be reduced. Thus, the recording time for the reference data set is lower and the imaging data for the reference data set can be selected independently of the parameters for the generation of the imaging data set. Preferably, the imaging data set is segmented into a plurality of imaging areas within a plurality of imaging layers, wherein the reference data set is captured for each of the plurality of imaging layers. If the imaging for the production of the MR image data is a multi-layer sequence, then the reference data record for this layer can be recorded before or after the acquisition of the actual layer.

For example, it is possible that an RF excitation pulse used to generate the spin-echo imaging data set has a different slice profile than an RF excitation pulse used in the creation of the reference data set. Likewise, the slice profile of the refocusing pulses used in the creation of the imaging data set may be different than the RF refocusing pulses used in the creation of the reference data set.

Furthermore, according to the invention, in the generation of the reference data set, RF refocusing pulses are used after a first RF refocusing pulse to generate the spin echoes which have a smaller tilt angle than the corresponding RF refocusing pulses after the first refocusing pulse used in the generation of the imaging data set. The tilt angle of the RF refocusing pulses after the first RF refocusing pulse may be between 20 ° and 120 ° in the reference data set, preferably between 20 ° and 70 °. Due to the lower tilt angle, the total recording time can be reduced since the time required for the irradiation of the RF pulses can be reduced. Furthermore, the incident energy can be reduced by lower tilt angles.

Likewise, it is possible to use RF excitation pulses and refocusing pulses in the generation of the reference data set whose spectral bandwidth is independent of the spectral bandwidth of the RF excitation pulses and refocusing pulses used in the generation of the imaging data set. For example, it is possible to select the spectral bandwidth at the RF excitation pulse and the RF refocusing pulse in the reference data set equal. This is usually not the case when creating the imaging dataset.

Furthermore, it is possible to choose the RF pulses used in the preparation of the reference data set such that they have an average reference time duration that is at least a factor of 0.8 shorter than an average time duration of the RF pulses that are generated at the time of creation of the imaging data set.

The above possibilities make it possible to record the reference data set in a very short time. This reference data set also has a good illumination and homogeneous distribution of the signal over the entire image data space. This makes it possible to calculate precisely the weighting matrix and the coil sensitivities for the individual receiving coils. By separating the acquisition of the raw data set and the imaging data set, more suitable raw data sets can be created for the calibration of the parallel imaging since no consideration has to be given to the parameters of the imaging data set used.

For example, the imaging dataset may be an imaging sequence of a turbo spin echo Dixon sequence. In such a sequence, spin echoes from different tissue components have at least once the same phase position and once an opposite phase position. In the prior art, it has been customary to interleave the reference data set with the actual imaging data set. In particular, in a turbo spin echo Dixon sequence, this led to unsatisfactory results in the preparation of the reference data set, as for the Fulfilling the condition for opposite phase positions of the distance of the refocusing pulses must be increased. If the reference data set is obtained from the data with the same phase position, the physiological effects described above are increasingly noticeable. However, when detected from the opposite phase data, the recording timings are not centered between two pulses, thereby providing additional dephasing of the spins in the peripheral regions of the FOV, which adversely affects calibration as described above.

The invention further relates to the associated MR system for generating the MR image data using the reference data set and the weighting matrix. The MR system has a control unit and a memory unit, wherein the memory unit stores executable control information from the control unit, wherein the MR system is designed to carry out the above-described and subsequently described steps in the execution of the control information in the control unit. In addition, a computer program product is provided with program means, wherein the computer program product can be loaded directly into a memory unit of a programmable control unit of the MR system in order to carry out the steps of the method described above, when the program means are executed in the control device. An electronically readable data carrier is provided with electronically readable control information stored thereon, which are equipped in such a way that when using the data carrier in a control unit of the MR system they perform the method described above.

The features described above, as well as the features described below, can be used not only in the correspondingly explicitly stated combinations, but also in other combinations, unless explicitly stated otherwise. Furthermore, the various features described may be used individually.

• 1 schematisch eine MR-Anlage, mit der erfindungsgemäß bei der Verwendung paralleler Bildgebungstechnik Artefakte vermindert werden können bei der Bildberechnung durch eine verbesserte Gewichtungsmatrix, die auf Grundlage eines Referenzdatensatzes berechnet wird,
• 2 schematisch ein Beispiel eines MR-Bildes, bei dem in einem Randbereich ein Artefakt auftritt, das mit der MR-Anlage von 1 vermieden werden kann,
• 3 schematisch ein Sequenzdiagramm zur Erstellung des Referenzdatensatzes, mit dem die Artefakte von 2 verringert werden,
• 4 ein Beispiel eines Flussdiagramms mit den Schritten zur verbesserten Aufnahme eines Referenzdatensatzes, mit dem eine verbesserte Berechnung der Gewichtungsmatrix möglich ist.

The invention will be explained in more detail with reference to the accompanying drawings. Hereby show:

• 1 schematically an MR system with which according to the invention when using parallel imaging technique artifacts can be reduced in the image calculation by an improved weighting matrix, which is calculated on the basis of a reference data set,
• 2 schematically shows an example of an MR image in which an artifact occurs in an edge region, which with the MR system of 1 can be avoided
• 3 schematically a sequence diagram for creating the reference data set, with the artifacts of 2 be reduced,
• 4 an example of a flowchart with the steps for improved recording of a reference data set, with which an improved calculation of the weighting matrix is possible.

1 1 schematically shows an MR system with which, according to the invention, artifacts can be reduced in MR images that are caused by a non-sufficient reference data set used to calculate a weighting matrix, the so-called kernel, in parallel imaging. The magnetic resonance system has a magnet 10 for generating a polarization field B0, wherein one on a couch 11 arranged examination person the examination object 12 represents that in an isocenter Z0 the magnet is moved to record there location-coded magnetic resonance signals of the examination subject. As will be explained in more detail below, the receiving volume usually has a first receiving area with a higher inhomogeneity of MR components and a second receiving area in which the homogeneity of the respective MR components is greater than in the first subarea. In this case, the polarization field B0 can be used as the MR component and / or the linearity of the magnetic field gradients, which are generated by gradient coils (not shown) for spatial coding. By radiating high frequency pulses and switching magnetic field gradients, the magnetization generated by the polarization field B0 can be disturbed by deflection of the nuclear spins from the equilibrium position and the return to the equilibrium position in receiving coils 5 to 8th induced currents can be converted into magnetic resonance signals. The general mode of operation for generating MR images under the detection of magnetic resonance signals, in particular in parallel imaging using a weighting matrix for generating unrecorded raw data points, is known to the person skilled in the art, so that a detailed description of this is dispensed with.

The magnetic resonance system furthermore has a control unit 13 which is used to control the MR system. The control unit 13 has a gradient control 14 for controlling and switching the magnetic field gradients and an RF control unit 15 for generating and controlling the RF pulses for the deflection of the nuclear spins from the equilibrium position. The RF unit may be a multi-channel RF unit that generates RF pulses in multiple independent channels. In a storage unit 16 For example, the imaging sequence necessary for the acquisition of the MR images can be stored as well as all others Control information necessary to carry out the invention. A picture sequence control 17 controls the image acquisition and thus, depending on the selected imaging sequence, the sequence of magnetic field gradients, the RF pulses and the reception intervals of the MR signals. This controls the image sequence control 17 also the gradient control 14 and the RF control unit 15 , In a computing unit 20 MR images can be calculated on a display 18 can be displayed. An operator can via an input unit 19 control the MR system. Among other things, the arithmetic unit can be used to calculate a weighting matrix or the so-called kernel, such as the GRAPPA kernel, which is calculated based on a reference data set.

In 2 is schematically an MR image 22 represented, which is an MR image of a first examination object 23 and a second examination object 24 shows. The examination object 24 is positioned in the phase encoding direction at the edge of the MR system, ie at the edge where the homogeneity of the polarization field B0 is lower and at which the linearity of the magnetic field gradients is lower than in the region in which the object 23 is arranged. At the in 2 In the embodiment shown, it is assumed that the first examination subject 23 is located substantially centrally in the isocenter Z0 of the magnet. In this second section 25 there is a first homogeneity of the polarization field B0 and the magnetic field gradients during image acquisition. Even after adaptation of all system components to the examination object, the so-called shimming, is a first subarea 26a and 26b present in which the magnetic field homogeneity and the gradient linearity are lower than in the second sub-range. This first subrange can be defined by the fact that the relationship between the physical location in the MR system and the Larmor frequency of the spins differs by 5%, 7% or 10% from the nominal value, which is a linear relationship between location and Larmor frequency when gradients are switched and when the gradient is off is a constant value. Due to the increased inhomogeneity in the partial area 26a on 26b, here in the subsection of the 26b it comes for the object under investigation 24 to increase tissue placement in a single pixel, with the phase encoding direction as in 2 shown perpendicular to the division of the receiving volume in the sub-areas. It can become artifacts 27 resulting in a high signal intensity due to the fact that several pixels are placed in the reconstruction on a single point, ie a single pixel in the MR image. Incorrect calibration of parallel imaging due to high signal intensity can cause these artifacts 27 at least partially as an additional artifact 27b in the visual field area 25 fold in and there the object 23 overlay, which can complicate the diagnosis. Such artifacts 27b can be reduced or prevented when calculating the weighting matrix of the parallel imaging as explained below.

In the following, it will be explained in detail how data can be provided by a separately recorded reference data set with a low resolution and the use of a multi-spin echo sequence for the generation of the reference data set, with which a good kernel or a good weighting matrix can then be calculated is used to calculate missing raw data points in parallel imaging. The MR images then created have fewer artifacts due to the improved reference data set.

The imaging sequence for creating the actual MR image data may be a Turbo Spin Echo Dixon imaging sequence. In this Dixon technique, the tissue components recorded in the examination subject have once the same phase position and once the opposite phase position. The MR images are usually taken in several segments, so that only a part of the raw data space is recorded per segment. Furthermore, the imaging sequence for creating the actual MR images is a multilayer sequence. For each layer of the multilayer sequence, the in 3 shown imaging sequence are used to generate the reference data set.

The sequence has an RF excitation pulse 30 and several refocusing pulses 31 . 32 and 33 on. In particular, the refocusing pulses 32 and 33 and all subsequent refocusing pulses may be chosen such that very small refocusing angles are used, for example between 20 ° and 120 °, preferably between 20 ° and 80 °, more preferably between 20 ° and 60 °. At the same time as the excitation pulses, in the slice selection direction, the Z direction of 3 , the slice selection gradients 40 . 40a . 41 and 42 switched, as they are commonly used in spin-echo sequences. The gradients occur in the phase coding direction 38 and 39 and other gradients of varying magnitude for phase encoding and the signal echoes generated by the refocusing pulses become during the readout gradients 34 . 35 - 37 read. This is schematically through the ADC 43 . 44 and 45 shown. The imaging sequence for generating the reference data set here differs from the imaging sequence that is used to create the imaging data set. So it is possible that the excitation pulse 30 and the refocusing pulses 31 . 32 or 33 have the same bandwidth. This is usually not the case with the RF pulses of the imaging sequence for creating the imaging data set. Due to the higher bandwidth and the lower flip angle, shorter time periods can be used for the irradiation of the excitation and refocusing pulses. For example, the average time period used for the excitation pulses and refocusing pulses is a factor of 0.6; 0.7 or 0.8, less than the corresponding averaged period of time that occurs in the excitation pulses and refocusing pulses of the imaging sequence for creating the imaging data set. In addition, the resolution in the readout direction is 4-10 times lower than when creating the imaging data set. Furthermore, other layer profiles may be used in the RF excitation pulses and refocusing pulses than in the associated RF excitation pulses and refocusing pulses to produce the imaging data set. As rectangular a layer profile as possible is not necessary in the reference data set, so that in turn the total recording time can be shortened by using other layer profiles which place less weight on a rectangular layer profile. The above measures reduce the sensitivity of the reference dataset to folding artifacts from the edge of the FOV.

The slice position for generating the MR images and thus the reference data set is preferably in a coronal or sagittal direction, where inhomogeneities occur at the edges due to the geometry of the MR system. Of course, however, the invention can also be used in transversal layer positioning.

Through the reference data set described above, this total can be recorded in a very short period of time, so that the total time duration for the acquisition of the imaging data set is also reduced. The data represented by the reference data imaging sequence of 3 generated reference data have a high homogeneity in the image data space. This makes it possible to use this reference dataset to calculate the weight kernel more accurately than with a reference dataset built into the dataset, which is used to subsequently reconstruct the raw data points that are not acquired in the parallel imaging that is taken using GRAPPA or SENSE techniques. How the weighting matrix is calculated from the reference data record is known to the person skilled in the art and will not be explained in detail here. The RF pulses used in the preparation of the reference data set further have the advantage that they bring very low RF exposure to the subject with it.

In 4 summarizes the steps that can be used to calculate missing raw data points in parallel imaging, with the reference data set as in 3 is discussed for creating a homogeneous reference data set, which forms a good basis for calculating the weighting matrix for the determination of the missing raw data points. The procedure starts in one step S51 and in one step S52 the reference data record is recorded as described above 3 was explained. Here, this reference data set is recorded for each imaging layer of an imaging data set. The recording of the individual layers preferably takes place in such a way that first the data for a first layer are completely recorded. This is followed by the complete recording of the data of a second shift. The second layer can be directly adjacent to the first layer or lie in another position in the layer stack, for example, in third place. Once all layers of the layer stack have been recorded, the weighting matrix for each layer can then be calculated, which then forms the basis for calculating the missing raw data points (step 52 ). In one step S53 then the imaging record is recorded. The recording of the imaging data set is usually segmented. Here, first of all, a first set of k-space lines is recorded from a first layer, for example every eighth k-space line of the raw data matrix. Subsequently, every eighth k-space line of a second layer is recorded, and so on. If this first number of k-space lines was recorded for each layer, a second pass of the entire layer stack takes place after the repetition time TR. Here, a second set of k-space line of the raw data matrix is recorded for each layer. The process is repeated until all necessary k-space lines have been recorded. Subsequently, in step S54 the missing raw data points using the weighting matrix and using the in step S53 recorded raw data points determined. Continue to be in step S55 reconstructs the MR images of the imaging data set. The reconstructed MR images have lower artifacts and overall, the recording time is reduced since the recording time for creating the reference data set can be reduced.

Claims (13)

Method for generating MR image data of an examination object with parallel imaging, comprising the following steps: - recording a reference data record of the examination object in which at least a subsection of the associated raw data space is completely filled with raw data according to the Nyquist condition, Taking an imaging data set for the generation of the MR image data on the basis of spin-echo-based signals, in which the associated raw data space is not completely recorded according to the Nyquist condition, - calculating a weighting matrix based on the reference data set with that in the parallel imaging not collecting raw data points of the imaging data set, calculating the unrecorded raw data points of the imaging data set using the calculated weighting matrix, wherein the reference data set is taken with a spin echo based reference imaging sequence without segmentation within an imaging layer and without temporal overlap with the acquisition of the imaging data set and in the readout direction, a resolution is used which is at least a factor of 4 lower than when recording the imaging data set, wherein in the preparation of the reference data set RF refocusing pulses after a first RF refocusing pulse for generating from the spin echoes having a smaller tilt angle than the RF refocusing pulses after the first refocusing pulse used in the preparation of the imaging data set. Method according to Claim 1 characterized in that the imaging data set is captured in a plurality of imaging layers with a segmented imaging method, wherein the reference data set is acquired for each of the plurality of imaging layers. Method according to one of the preceding claims, characterized in that an RF excitation pulse used in the preparation of the imaging data set for generating the spin echo has a different slice profile than an RF excitation pulse, which is used in the preparation of the reference data set. Method according to one of the preceding claims, characterized in that, in the generation of the imaging data set, RF refocusing pulses are used to generate the spin echoes which have a different slice profile than the slice profile used in the RF refocusing pulses to generate the reference data set , Method according to Claim 1 wherein the flip angles of the RF refocusing pulses after a first RF refocusing pulse used in the preparation of the reference data set are between 20 ° and 120 °, preferably between 20 ° and 70 °. Method according to one of the preceding claims, characterized in that are used in the preparation of the reference data set RF excitation and refocusing, whose spectral bandwidth is selected independently of the spectral bandwidth of the RF excitation and Refokussierungspulse that is used in the preparation of the imaging data set , Method according to Claim 6 , characterized in that the spectral bandwidth at the RF excitation pulse and the RF refocusing pulses is the same in the reference data set. Method according to one of the preceding claims, characterized in that the RF pulses used in the preparation of the reference data set have an average reference time duration that is shorter by at least a factor of 0.8, preferably 0.6, than an average time duration of the RF pulses, used in the creation of the imaging data set. Method according to one of the preceding claims, characterized in that the imaging sequence used to create the imaging data set is a turbo-spin echo Dixon sequence in which spin echoes of different tissue components of the examination subject have the same phase position at least once and at least once have opposite phase position. An MR system (10), which is designed to generate MR image data of a subject under examination with parallel imaging, wherein the MR system has a control unit (20) and a memory unit (21), wherein the memory unit stores executable control information from the control unit, the MR system is designed to carry out the following steps when carrying out the control information in the control unit: - capturing a reference data record of the examination object in which at least a subsection of the associated raw data space is completely filled with raw data according to the Nyquist condition, - capturing an imaging data set for the Generation of the MR image data on the basis of spin-echo-based signals, in which the associated raw data space is not completely recorded according to the Nyquist condition, - calculating a weighting matrix from the reference data set, with the raw data points of the raw data not taken in the parallel imaging Imaging data set to be determined, - calculating the unrecorded raw data points of the imaging data set using the calculated weighting matrix, wherein the reference data set with a spin-echo-based Reference imaging sequence is recorded without segmentation within an imaging layer and without temporal overlap with the recording of the imaging data set and in the readout direction, a resolution is at least by a factor of 4 lower than when recording the imaging data set, wherein in the preparation of the reference data set RF Refocusing pulses are used after a first RF refocusing pulse to generate the spin echoes having a smaller tilt angle than the RF refocusing pulses after the first refocusing pulse used in the preparation of the imaging data set. MR system after Claim 10 , characterized in that the MR system is formed when executing the control information in the control unit, a method according to one of Claims 2 to 9 perform. Computer program product, which comprises program means and can be loaded directly into a memory unit of a programmable control unit of an MR system in order to perform all the steps of the method according to one of the Claims 1 to 9 execute when the program means are executed in the control device. Electronically readable data carrier with electronically readable control information stored thereon, which are designed such that when using the data carrier in a control unit of an MR system, the method according to one of the Claims 1 to 9 carry out.

Images